基于LDA模型的元宇宙主题挖掘与演化

Metaverse Topic Mining and Evolution Based on LDA Model

检索WoS核心数据库关于元宇宙文献(保留文献的摘要部分),依靠Python进行LDA(潜在狄利克雷分配)模型构建,通过困惑度以及主题一致性确定最佳主题数,并抽取文本下词频最高的11个主题,对元宇宙相关文献进行主题抽取与挖掘,运用WordCloud技术、桑基图(主题相似度)进行主题演化分析,从而了解元宇宙发展状态以及演化趋势。结果表明:目前关于元宇宙的研究主要在“元宇宙技术”和“元宇宙应用”方面,词云分析中关键词“tech-nolog”以及“virtual”较为突出;从主题热度分析来看,主题4(教育元宇宙研究)、主题7(元宇宙技术研究)、主题8(元宇宙应用研究)是目前元宇宙研究的热点;从主题演化分析来看,第1阶段(2018—2022年)中主题2元宇宙模型研究和主题8元宇宙技术研究表现出较强的演化能力,第2阶段(2023—2024年)的主题4元宇宙体验研究和主题8元宇宙场景构建研究演化能力强,且与相邻主题相似性高。

The metaverse literatures(retaining the abstract part of the literature)in the WoS core database were retrieved,the LDA(latent Dirichlet allocation)model was built based on Python.The optimal number of topics through perplexity and topic consistency are determined,and the 11 topics with the highest word frequency under the text were extracted.The topics related to metaverse were extracted and mined,and WordCloud technology and Sankey diagram(topic similarity)were used to analyze the topic evolution,so as to understand the development status and evolution trend of metaverse.The results show that the current research on metaverse is mainly in the aspects of“metaverse technology”and“metaverse application”,and the keywords“technolog”and“virtual”are more prominent in the word cloud analysis.From the perspective of topic heat analysis,Topic 4(education metaverse research),Topic 7(metaverse technology research),and Topic 8(metaverse application research)are the hot spots of metaverse research.From the perspective of topic evolution analysis,in the first phase(from 2018 to 2022),topic 2 metaverse model research and topic 8 metaverse technology research showed strong evolutionary capabilities,while in the second phase(from 2023 to 2024),topic 4 metaverse experience research and topic 8 metaverse scene construction research showed strong evolutionary capabilities and high similarity with adjacent topics.